XLRI Jamshedpur

Synthetic CDO

Himadri Singha (019)

Kumar Vikram (024)
Hozefa Bharmal (078) Group 3
Ritu Agarwal (102)
Subhadip Das (110)
Nikhil Uppal (092)
Basics of Synthetic CDO

This product was introduced where Credit Risk Transfer
was more important

Credit Risk is transferred by Originator to the Investors by

means of CD instruments

Risk transfer is undertaken by an SPV

Originator is the “Protection Buyer” and Investors are

“Protection Seller”

Main purpose is to mitigate risk without any asset transfer.

Cash CDO Vs Synthetic CDO
Cash CDO Synthetic CDO

Involve a portfolio of cash
Do not own cash assets
assets (corporate bonds)

These CDOs gain

Ownership of assets is exposure only to the
transferred to SPV, assets through CDS.
issuing the tranches

SPV doesn’t bear the

SPV bears the operational risk
operational risk
Synthetic CDO Structure
Originator Default
Payment P&I Investors
Asset
SPV Coupon
Payment Senior
(Protection Seller)
Asset
Proceeds Mezzanine
Asset
CDS
Premium Equity

Trustee

High Quality
Asset
Waterfall Diagram
CDS
Premium

Default Payment

Low Risk
Senior Tranche
Low Yield

Mezzanine Tranche
High Risk
Equity
High Yield
Types of Synthetic CDO

 Unfunded Synthetic CDO

 Protection seller’s payment obligation is not paid upfront
 Investors are ultimate protection seller.

 Funded Synthetic CDO

 Protection seller’s payment obligation is paid upfront through issuing
CLN

 Proceeds from CLN are invested in Risk Free assets

 Partially Funded Synthetic CDO

Unfunded Synthetic CDO
Protection buyer enters into a
CDS with SPV, which in turn,
enters into a CDS with
investors, the ultimate
protection seller
 Funded Synthetic CDO
Interest payment equal to the yield on
high quality asset + CDS Premium

Originator Default
Payment Investors
Coupon
Asset LIBOR +
SPV X bps Senior
(Protection Seller)
Asset
Proceeds Mezzanine
Asset
CDS From CLN
Premium Equity

This is done to “delink” the Trustee

credit ratings of the notes
from the rating of the Notes equal to 100% of
originator. the value of the ref pool
High Quality
Else downgrade of the
Asset of assets are issued
originator would downgrade
the issued notes.
Partially Funded Synthetic CDO
SST does not pay purchase
Perceived risk is less price. Rather SST receives
payments as protection seller
5-10% default risk
and is liable to pay the originator
if the underlying assets suffer a
loss above specified level.
Originator CDS Super
Premium
Senior
Unfunded
Protection
Asset Tranche
Pay if default Investors
Asset
CDS Coupon
Asset Premium SPV Libor+ Senior
(Protection
Asset Seller)
X bps

Asset Funded Proceeds Mezzanine

Pay if default From CLN
Tranche
Asset
Trustee Equity

High Quality
Asset
A typical funding structure

Cash CDO Synthetic CDO

% of % of
Grade Tranche Size Tranche Size
Portfolio Portfolio

Super Senior 43,25,00,000 86.50%

Aaa 43,95,00,000 87.90% 2,00,00,000 4.00%

Aa2 11.500,000 2.30% 12.500,000 2.50%
Baa2 14.000,000 2.80% 15.000,000 3.00%
Equity 3,50,00,000 7.00% 2,00,00,000 4.00%
Total 50,00,00,000 100.00% 50,00,00,000 100.00%
Motivation
Typically the reference assets are not actually removed from the sponsoring
firm’s balance sheet. For this reason:

Synthetic CDOs are easier to execute than cash structures
 the legal documentation and other administrative requirements are less
burdensome

Synthetic CDO ensures transfer of credit risk of assets not suited for
conventional securitization, while the actual assets are retained on the
balance sheet.
 For example, Bank guarantees, Letter of Credit etc.

A more efficient way of Credit risk mitigation

Originator does not have to reduce book size as BS remains unchanged

The super senior tranche, which prices well below a typical AAA tranche
and which makes up more than 80% of the synthetic CDO, is a major driver
of the economics of the synthetic CDO
Motivation
Senior Tranche Super Senior
(86%) Swap(92.5%)
Cash LIBOR+40bp Synthetic 15 bp
Flow CDO Mezzanine (6%)
CDO
Senior (2.5%)
LIBOR+70bp LIBOR+30bp
1 billion 1 billion
Mezzanine (6%) Mezzanine
dollar dollar (3%)
LIBOR+165bp
Reference Reference LIBOR+165bp
Portfolio Equity (2%) Portfolio Equity (2%)

That means if CDO manager can reinvest in collateral pool risk free
asset at, say, (LIBOR-5 bp), it is able to gain from a savings of 20 bp on
each 100 dollar if structure is unfunded

A Considerable Gain
Structure of a CDO Tranche
Traditionally, a collateralized debt obligation pool is divided into three
tranches; wherein each tranche behaves as a separate CDO, enabling
the CDO originators to attract multiple investors having varying risk
preferences
1. Senior Tranche or Senior Debt: This is typically highly rated, since
it is ranked on top in terms of priority of payments. However, the
interest rate on investments in this tranche is the lowest due to the
lower risk that accompanies them
2. Mezzanine Tranche: This tranche has moderate returns and
moderate risk
3. Equity Tranche: Investment in this tranche yields the highest
interest rate. This high rate is offered to counter the higher risk on this
tranche. Equity tranche investors are the first to lose funds when loans
in the pool are not repaid
Single Tranche CDO
Also known as ‘tailor made CDOs’, they are customized to
meet the individual investor needs with respect to:

 Portfolio Size
 Asset Classes
 Portfolio diversity and rating
 Portfolio geographical and industrial variation
 Portfolio term to maturity
 Type of collaterals used
 Subordination level
Single Tranche CDO
 Single-tranche CDOs represent the vast majority of all new synthetic
CDO issuances.
 The CDO manager sells only a single tranche – usually at the
mezzanine level – of the capital structure to an investor instead of
selling all the tranches at the same time
 The Single Tranche CDO can be issued either directly by the Banks
or via SPVs

Advantages of Single Tranche:

 Single tranche is tailored to the specific investor’s needs

 It is not necessary for the CDO manager to find investors across the
entire capital structure simultaneously
Risk Associated with Synthetic CDO
 Risk of the underlying asset
 Due to the absence of a true sale of the underlying assets,
synthetic CDOs involve the credit risk inherent in the underlying
assets. These assets could be bonds, ABS, MBS, loans etc. The
risk of these assets is generally measured using their credit
rating, historical performances and any other asset specific
information.
 Legal issues associated with the CDO definition
 As there is a conflict of interest between the protection buyer and
the protection seller on the occurrence of a credit event it is of
prime importance that the “trigger events” be clearly defined.
 Counterparty credit risk
 There is a risk of the counterparty’s inability to pay in case of
credit default
Confidentiality & Tax Issues
Confidentiality
Generally the Protection Buyer cannot share the names of the
Reference Entity with the Protection Seller due to issues of
confidentiality. In order to counter this situation one can nevertheless

 Define general eligibility criteria with which the Reference Obligations

 and Reference Portfolio must comply,
 Appoint the Protection Buyer itself as calculation agent (who determines
whether or not a Credit Event has occurred) and
 Give a supervising role to the Protection Buyer’s external auditors.

Tax Issues
Since the title of the reference Obligations are not transferred to the
Protection Seller, taxation is not a major consideration in the case of a
Synthetic CDO
Moody’s Ratings Framework

Moody's rating on each rated note represents the expected loss
on the note, which is the difference between the present value
of the expected payments on the note and the present value of
the promised payments under the note, expressed as a
percentage of the present value of the promise

To evaluate the expected loss, Moody’s incorporates both

quantitative and qualitative analysis

Moody's expected loss models capture the quantifiable risks

while a legal review of the transaction seeks to ensure that non-
quantifiable risks are mitigated through documentation
provisions
Quantitative Analyses

The primary source of risk in a synthetic CDO comes from the
reference pool

Moody’s uses the quantitative analysis to assess the risks

stemming from the reference pool

The premium payments are excluded from the scope of the

quantitative analysis because the promised premium is large
enough to ensure coverage of the interest payments on the
CDO

There are two primary methods to model a default risk:

 Binomial Expansion Modeling
 Multiple Binomial Modeling
Binomial Expansion Modeling

Primarily used for a pool of homogeneous assets

A model portfolio is created which contains a pool of N diversity

bonds

Each diversity bond is assumed to have identical characteristics

in terms of par/notional amount, rating, average life, spread and
recovery, and is uncorrelated with every other diversity bond in
the pool

The number of diversity bonds in the portfolio is equivalent to

Moody's diversity score
Binomial Expansion Modeling

The losses stemming from the default of each additional diversity
bond in the model portfolio going from zero diversity bond
defaults to N diversity bond defaults is calculated and a
probability assigned to each default scenario

Calculating this probability-weighted loss for each CDO tranche

generates the expected loss
Multiple Binomial Modeling

An extension of the Double Binomial Method, used in cases where the
underlying portfolio assets exhibit heterogeneous characteristics -
such as having a clear delineation between low rated and highly rated
assets

Moody’s divides a pool of reference entities/credits into the most

appropriate number of sub-pools and models the default behavior of
each pool with a separate binomial analysis

Each diversity bond is assumed to have identical characteristics in

terms of par/notional amount, rating, average life, spread and
recovery, and is uncorrelated with every other diversity bond in the
pool
Multiple Binomial Modeling
The mathematical expression for the multiple binomial-based
expected loss used by Moody’s is as below:
Multiple Binomial Modeling
Factors which warrant the use of the Multiple Binomial Method to quantify
the inherent risks are:

 Portfolio Characteristics

Most synthetic CDOs have reference entities/credits whose ratings

can vary greatly (typically Aaa down to Baa3 or even Ba3), for a 5-
year synthetic CDO, Moody's idealized default probability can vary
from as little as 0.003% for a Aaa credit to 3.05% for a Baa3 credit
and 11.86% for a Ba3 credit
Multiple Binomial Modeling
 Capital Structure

Most synthetic CDOs are highly leveraged and are thus sensitive to
fewer defaults than cash flow CDOs .Hence only a small amount of
subordination is necessary to support high ratings. This thin
subordination combined with the relatively small sizes of the rated
tranches generally requires more precision in the calculation of the tail
probability of the loss distribution.

 Structural Features, or Lack Thereof

Many synthetic CDOs do not have the ability to generate any excess
spread that may be used to offset losses in the reference pool. Hence,
it is even more important to capture the correct loss distribution when
analysing the expected loss of a CDO tranche
Qualitative Analysis

In case risks inherent in a synthetic CDO are not or cannot be modeled
quantitatively, they would be addressed through the legal
documentation, and hence the importance of Qualitative Analysis

The important aspects of the qualitative analysis unique to synthetic

CDOs can be grouped into three main categories:

 Trading guidelines for managed synthetic CDOs

 Credit event definitions and their effects on the modeled default
probabilities
 Structural features such as valuation procedures and settlement
mechanisms that affect recovery rate assumptions.
NIG for Synthetic CDO Pricing
 Normal Inverse Gaussian Distribution for Synthetic CDO pricing is an
extension of the popular Large Homogeneous Portfolio (LHP),
approach to CDO pricing

 LHP assumes a flat default correlation structure over the reference

credit portfolio and models defaults using a 1-factor Gaussian copula

 This model leads to an implied correlation skew, as it fails to fit the

prices of different CDO tranches simultaneously

 This is explained by the lack of tail dependence in the Gaussian

copula and a Student t-distribution is proposed

 However, the t-distribution leads to an increase in computation time

and therefore the NIG is proposed
NIG for Synthetic CDO Pricing
 Normal Inverse Gaussian Distribution is a special case of the
generalized hyperbolic distribution

 They are flexible four parameter distribution family that can produce fat
tails and skewness
Properties of NIG
 Normal Inverse Gaussian Distribution is a mixture of the normal and
the inverse Gaussian distributions

 They are flexible four parameter distribution family that can produce
fat tails and skewness

 A non-negative random variable Y has an Inverse Gaussian

distribution with parameters:

Hence
Properties of NIG
 A random variable X follows a Normal Inverse Gaussian Distribution
with parameters

 They density and probability functions are thus:

Properties of NIG
 The main properties of the NIG distribution class are the scaling
property:

 And the closure under convolution for independent random variables X

and Y:
Derivation of Pricing formula using
NIG:
 Since M does not depend on a, we set:

 The random variable,

 Is NIG distributed and its parameters are:

Derivation of Pricing formula using
NIG:
 Thereafter the 3rd and 4th parameters are restricted to standardize
the distributions of both the factors:

 With
Derivation of Pricing formula using
NIG:
 Starting with:

 Then applying the scaling property we get:

 Thereafter applying the convolution property to

Derivation of Pricing formula using
NIG:
 Finally, we get:

 The above is the expression for the NIG distribution function and
the density
Advantages
 Time to market is less compared to a cash deal, with average execution
time typically varies from six to eight weeks based on the structure
compared to three to four months for an equivalent cash deal
 Leads to lower transaction cost as SPV setup cost can be avoided
 Use of credit derivatives offer greater flexibility for risk requirement
 Cost of buying protection is lower and credit protection price is below the
note liability
 Range of reference asset is wider and typically includes bank guarantee,
derivative instruments
 Clients whose loans need not be sold off from the sponsoring agent’s
B/S can be better handled and leads to improved customer relationship
 Credit default swap is cheaper than the underlying cash bond for many
reference names
Factors to consider during analysis

Three key factors are being considered by analystP

 Default probabilities and cumulative default rates

 Default correlations
 Recovery rates

Default Probability rates

 A number of methods are used to estimate default probabilities like
individual credit ratings and historical probability rates
 Common method is to use average rating of the reference portfolio which
consists of 150 or more reference names
 Rating agencies like Moody’s provide data on the default rates for
different ratings as an average class
Correlations
 Correlations among underlying assets pool is taken into
consideration during analysis

 Because correlation is unobservable, differences of opinion

among market participants as to the correct default
correlation creates trading opportunities

 Diversity score of a CDO plays a part in calculating the

precise correlation value which is used to map the
underlying CDO portfolio into a hypothetical portfolio
consisting of homogeneous assets

 It represents the number of uncorrelated bonds with

identical par value and with the same default probability
Recovery rates
 Generally analyst constructs a database of recovery rates by
industry types and credit ratings used by different agencies

 However, for synthetic CDOs with credit default swap as assets in

the portfolio, this factor needs to be ignored

 Analyst performs simulation model to generate scenarios of

default and expected returns

 All variables like the number of defaults swap to maturity, recovery

rates and timing of defaults etc. are considered as random and
thus modeled using stochastic process

 However, actual recovery rates might differ based on the

macroeconomic factors